Probing the Early Events Associated with Liquid ... - ACS Publications

Jul 5, 2011 - César A. López, Giovanni Bellesia, Antonio Redondo, Paul Langan, Shishir P. S. Chundawat, Bruce E. Dale, Siewert J. Marrink, and S...
0 downloads 0 Views 6MB Size
ARTICLE pubs.acs.org/JPCB

Probing the Early Events Associated with Liquid Ammonia Pretreatment of Native Crystalline Cellulose Giovanni Bellesia,† Shishir P. S. Chundawat,‡,^ Paul Langan,§,r Bruce E. Dale,‡,^ and S. Gnanakaran*,|| †

T6 & CNLS, Los Alamos National Laboratory, Los Alamos, New Mexico 87545, United States Biomass Conversion Research Laboratory, Department of Chemical Engineering and Material Science,Michigan State University, Lansing, Michigan 48824, United States § Bioscience Division, Los Alamos National Laboratory, Los Alamos, New Mexico 87545, United States T6, Los Alamos National Laboratory, Los Alamos,New Mexico 87545, United States ^ DOE Great Lakes Bioenergy Research Center, East Lansing, Michigan 48824, United States

)



bS Supporting Information ABSTRACT: Various chemicals are being explored for catalyzing efficient lignocellulose deconstruction. In particular, when liquid ammonia is used to convert the naturally occurring cellulose crystalline phase Iβ, to cellulose IIII, the rearrangement of the hydrogen bond network in cellulose IIII results in enhanced hydrolysis yields. We use molecular dynamics simulations to analyze the interaction between a cellulose Iβ fibril and ammonia. Our simulations reveal that early structural changes in the fibril are driven by the rapid formation of an extended hydrogen bond network between the solvent-exposed surface chains and ammonia that precedes ammonia penetration into the fibril. The emergence of this hydrogen bond network causes relative shifting of the cellulose layers within the fibril that in turn leads to the formation of channels orthogonal to the (100) and (100) fibril surfaces. The channels allow ammonia molecules to penetrate into the cellulose fibril. These findings provide avenues for improving existing chemical pretreatments to make them more effective and economical.

’ INTRODUCTION Cellulose, a linear polymer composed of β(14) linked D-glucose units and a major component of lignocellulosic biomass, is synthesized at the plasma membrane of growing plants from arrays of cellulose synthase complexes1 as crystalline nanofibers whose unit cell parameters and space group are identified as cellulose Iβ. In a typical process of biomass conversion for biofuels production, thermochemical pretreatments are, first, used to improve enzyme accessibility of cellulose fibers which, in untreated biomass, are embedded in an amorphous matrix of hemicellulose and lignin. Then cellulose is hydrolyzed by enzymes to glucose that can eventually be converted to biofuels via microbial fermentation or chemical catalysis.2,3 However, intracrystalline hydrogen bonds and stacking interactions in cellulose Iβ fibers make crystalline cellulose recalcitrant to enzymatic hydrolysis, therefore hindering the efficiency of the celluloseto-glucose conversion process.24 One route to improve cellulose hydrolysis relies on the use of liquid ammonia to catalyze a structural crossover from cellulose Iβ to another crystalline cellulose allomorph named cellulose IIII. This structural crossover leads to rearrangement of both the intracrystalline hydrogen bonds and the stacking interactions between the glucose rings.5,6 Recent experiments have shown that such a rearrangement in cellulose IIII enhances the r 2011 American Chemical Society

enzymatic hydrolysis rates by up to 5-fold with respect to cellulose Iβ.7 In addition to the enhancement in cellulose digestibility, ammoniabased treatments (e.g., AFEX) have additional advantages over other pretreatment technologies in terms of both economic feasibility and environmental impact.8 X-ray and neutron diffraction structures for cellulose Iβ, cellulose IIII as well as for the transient ammoniacellulose I complex are known.6,9,10 The latter is an intermediate state where ammonia molecules are trapped in specific sites within the cellulose crystal and form hydrogen bonds with the neighboring cellulose chains.10 Cellulose IIII is generated from the ammoniacellulose I complex upon ammonia evaporation which prompts an alteration in the hydrogen bond network and a related relatively small rearrangement of the packed cellulose chains.6,9,10 While the experimental protocol for the ammoniabased conversion from cellulose Iβ to cellulose IIII is well established, the atomistic details of the structural crossover from cellulose Iβ to the ammoniacellulose I complex and to cellulose IIII are still unknown. In this study we use an extensive set of fully atomistic molecular dynamics (MD) simulations to analyze the Received: May 25, 2011 Revised: June 30, 2011 Published: July 05, 2011 9782

dx.doi.org/10.1021/jp2048844 | J. Phys. Chem. B 2011, 115, 9782–9788

The Journal of Physical Chemistry B early stages of the interaction between liquid ammonia and a cellulose Iβ fibril. We observe a sequence of events characterizing the early stages of the interaction between liquid ammonia and native crystalline cellulose. These events can be summarized as follows: (I) Formation of hydrogen bonding network between the ammonia molecules in the vicinity of the cellulose fibril and the cellulose chains on the surface of the fibril. (II) Consequential change in the hydroxymethyl group rotational state population from tg (the rotational state observed in cellulose Iβ) to gt and gg (the rotational states observed experimentally and in previous simulation studies, in the ammoniacellulose I complex, in cellulose IIII as well as in amorphous and fully solvated cellulose). The change in the hydroxymethyl rotational state population appears to start from the cellulose chains exposed to the ammonia solvent and then quickly propagate to the core of the fibril. (III) Structural crossover from cellulose Iβ to a new crystalline structure where the chain stacking arrangement in the plane perpendicular to the fibril main axis is similar to that observed in the ammoniacellulose I complex and cellulose IIII. (IV) Ammonia penetration appears to follow major fibril structural changes, suggesting that liquid ammonia and its interaction with the cellulose surface is a prerequisite for structural changes required for the conversion from cellulose Iβ to cellulose IIII.

’ MATERIALS AND METHODS X-ray and neutron diffraction coordinates5 were used to generate one “rhomboid” cellulose Iβ fibril composed by 30 octameric glucan (cellulose) chains. All MD simulations were performed under NPT conditions (isothermalisobaric ensemble), using the NAMD software,11 with the GLYCAM0612 force field. The following parametrization was used for the ammonia molecule:13,14 Ncharge = 1.026e, Hcharge = 0.342e, Kbond = 394.1 kcal/mol, r0 = 1.018 Å, Kangle = 41.3 kcal/mol, θ0 = 107.13 (deg), ANN = 9.44293233  105, ANH = 2.12601181  103, AHH = 1.39982777  101, BNN = 8.01323529  102, BNH = 2.09604198  10, BHH = 9.37598976  102. Units for the A and B Lennard-Jones coefficients are (kcal/mol) Å12 and (kcal/ mol) Å6, respectively. Results from a MD simulation of a liquid ammonia box show good agreement with both the experimental density and the experimental NN radial distribution function (data not shown). Further validation of the ammonia model comes from a recent neutron diffraction and molecular dynamics study where both the structure and the dynamics of an ammonia cellulose I infinite crystal were studied in detail.10 The fibril was solvated in a rectangular ammonia box, and each glucan chain was covalently connected to its periodic image along its main axis in order to mimic an infinitely long chain (fibril). A snapshot of the solvated cellulose Iβ fibril is given at the top of Figure 1. The initial ammonia coordinates were obtained from an equilibrated box of liquid ammonia. A Langevin thermostat and a NoseHoover Langevin barostat15,16 were used to control the temperature and the pressure, respectively. The damping coefficient for the Langevin integrator was set to b = 1.0 ps1, whereas for the NoseHoover Langevin barostat we applied an oscillation period of 200 fs and a damping period of 100 fs. The cutoff for the nonbonded interactions in the coordinate space was fixed at 10.0 Å. All of the simulations were performed under periodic boundary conditions, and the longrange electrostatic interactions were calculated by using the Ewald summation method

ARTICLE

with the particle mesh Ewald algorithm.17 The particle mesh Ewald accuracy was fixed at 106, the order of the interpolation functions on the grid was set to 4 (cubic), and the grid size was ∼1.0 Å. The time step was fixed in all simulations at 2.0 fs. The covalent bonds involving hydrogen atoms were constrained by means of the SHAKE algorithm.18 We considered more than 60 distinct MD simulations totaling 832 ns to capture nonequilibrium dynamics involving ammonia cellulose interaction. The simulation set #1 was carried out to analyze the early changes in the hydroxymethyl group rotational state as well as the cellulose and celluloseammonia hydrogen bonds networks. Here, the initial conformation was generated performing, first, a local optimization followed by a short thermal relaxation of the liquid ammonia with the cellulose fibril atoms kept fixed. This initial conformation was used to start 40 independent NPTMD simulations, 100 ps each. The pressure was fixed at P = 1.01325 bar, whereas two different temperatures were used (each for 20 MD simulations): T = 298 and 413 K.9 The simulation set #2 was carried out to probe the onset of the “macroscopic” structural changes in the fibril (changes in the mutual distances of the glucan chains within the fibril) as well as the ammonia molecules “penetration” into the fibril. The initial conformation for this set of simulations was generated performing, first, a local optimization followed by a short (∼1.0 ns) NPTMD simulation (P = 1.01325 bar), where the temperature was gradually increased from 100 to 298 K (no atoms were fixed). This initial conformation was used to start 20 independent NPTMD simulations. Of these 20 simulations, 10 were 10 ns long with T = 298 K, and 10 were 50 ns long with T = 413 K. Additional MD simulations (set #3) on ammoniasolvated cellulose fibrils were performed at T = 413 K on both a cellulose Iβ and a cellulose IIII fibril (76 and 62 ns, respectively), and at T = 450, 550, and 650 K (set #4) on a cellulose Iβ fibril (30 ns each). The temperature T = 413 K was chosen in accordance with the experimental temperature used by Wada et al. to obtain ammoniacellulose I complex from native cellulose Iβ.9 Higher temperatures (T = 450, 550, and 650 K) were chosen in order to observe a more extensive ammonia penetration within reasonably short simulation time lengths. Hydrogen bonds within the cellulose fibril and between cellulose and liquid ammonia were defined by a cutoff distance of 3.3 Å and a cutoff angle of 30.1921 The values of the dihedral angles χ(O5C5O6C6) and χ0 (C4C5O6C6) were employed to define the rotational state of the hydroxymethyl group. Three main rotational states were considered: (i) tg (transgauche), typically found in cellulose Iβ,5 (ii) gt (gauchetrans), dominant in cellulose IIII6 and in the ammonia cellulose I complex,9 and also observed in both amorphous and solvated cellulose22,23 and (iii) gg (gauchegauche), observed in both amorphous and solvated cellulose.22,23 The ammoniainduced structural changes were monitored via three parameters: the distances a0 and b0 , and the angle γ0 , all defined in Figure S1. In the cellulose Iβ structure, a0 = (a2 + b2 + 2ab cos(γ))1/2 and γ0 = 180  arcsin (sin(γ)a/(2a*)) where a, b0 = b and γ define the crystal unit in the plane perpendicular to the cellulose chain main axis,5 and a* = 0.5(a2 + b2  2ab cos(γ))1/2. In the ammoniacellulose I complex and in the cellulose IIII structures, a0 = 2a, b0 = b and γ0 = γ.6,9 In cellulose Iβ (our starting configuration), a0 = 10.7 Å and γ0 ∼ 140. a0 = 8.9 Å in both the ammoniacellulose I complex and cellulose IIII while γ0 = 92.7 and γ0 = 105 in the ammoniacellulose I complex and in cellulose IIII, respectively. 9783

dx.doi.org/10.1021/jp2048844 |J. Phys. Chem. B 2011, 115, 9782–9788

The Journal of Physical Chemistry B

ARTICLE

Figure 1. Top: starting configuration: ammoniasolvated cellulose Iβ fibril. Bottom: cellulose fibril compatible with ammoniacellulose I complex symmetry obtained after 76 ns at T = 413 K. Only partial ammonia penetration is observed. The black parallelograms indicate the respective crystal units in the ab plane.

’ RESULTS AND DISCUSSION At very early time scales, we monitor rapid formation of hydrogen bonding between NH3 and exposed chains of cellulose. The simulation set #1 that includes more than 40 separate simulations of 100 ps each was used to characterize these early events. At time t = 0, the “internal” cellulose hydrogen bond network was the one typical of cellulose Iβ5 and there were no hydrogen bonds between cellulose and the ammonia molecules. Then, in a few picoseconds, we observe a rapid increase in the

number of hydrogen bonds between ammonia molecules and solvent-exposed cellulose chains. This new hydrogen bond network involves mainly backbone oxygen atoms O2 and O3 as well as sidechain oxygen atom O6 in the hydroxymethyl group. The time evolution for the number of celluloseammonia hydrogen bonds considering O6, O3, and O2 as both donors and acceptors, and O5 as acceptor at T = 298 K is shown in Figure 2. The hydrogen bonding network profile is slightly lower at higher temperature (T = 413 K, Figure S2 in the Supporting 9784

dx.doi.org/10.1021/jp2048844 |J. Phys. Chem. B 2011, 115, 9782–9788

The Journal of Physical Chemistry B

ARTICLE

Figure 2. Early (100 ps) time evolution for the number of celluloseammonia hydrogen bonds, considering cellulose atoms O6, O3, and O2 as both donors (D, red) and acceptors (A, blue) and O5 as acceptor at T = 298 K. The plots are generated from values averaged over 20 trajectories. The errors vary between 0.5 and 1.5.

Information). The minor contribution of O4 as acceptor is not shown. The “internal” cellulose hydrogen bond network also undergoes substantial changes within the 100 ps MD simulation length, namely, the numbers of intrasheet hydrogen bonds (intra- and intermolecular) decrease while the number of intersheet hydrogen bonds increases7 (Figure S3 in the Supporting Information). The same set of simulations show a decrease in the hydroxymethyl group rotational state tg compensated by a corresponding increase in both gt and gg rotational states. At time t = 0, all of the hydroxymethyl groups in the cellulose fibril were in the tg state (tgfraction = 1.0). The subsequent observed shift in rotational states appears to be a consequence of the rapid formation of hydrogen bonds between ammonia molecules and solvent exposed cellulose chains. In Figure 3 we show representative plots of the time evolution (over t = 100 ps) of the tg, gt, and gg populations for solventexposed chains at T = 413 K and T = 298 K (A and B, respectively) as well as for cellulose chains in the crystalline core at T = 413 and 298 K (C and D, respectively). The decrease in the tg population is faster in the solventexposed chains than in the chains within the crystalline core of the fibril. The changes in the hydroxymethyl group rotational state occupancy are also generally faster at higher temperature (See Figure S4 in the Supporting Information). Along much longer time scales, we observe global structural changes that affect the crystalline structure of native cellulose. Structural rearrangements in nanosecond time scales were captured from the simulation set #2 that considered more than 20 separate

simulations of 50 ns each. The ammoniainduced structural changes were monitored via three parameters: the distances a0 and b0 and the angle γ0 (see the Supporting Information). At T = 298 K (see Figure 4, panels A and B), both a0 and γ0 evolve, over 10 ns, from their initial values typical of the cellulose Iβ crystalline structure to values consistent with both the ammoniacellulose I complex and cellulose IIII. Remarkably, no ammonia penetration is observed in any of the 10 ns trajectories at T = 298 K, confirming that early structural changes in our model fibril are mainly due to the effect of liquid ammonia as a bulk solvent surrounding the cellulose fibril. At T = 413 K, parameters a0 and γ0 reach a plateau value of ∼8.5 ( 0.4 Å and ∼114 ( 4, respectively, both after ∼3035 ns. Both of these crystalline parameters are consistent with a cellulose IIII structure. For the purpose of clarity, only the first 4 ns of the total 50 ns are shown in Figure 4, panels C and D. The crystal unit parameter b0 (average value is 7.6 ( 0.4 Å for both T = 298 and 413 K), however, does not change during the global structural change. We do not observe a complete structural relaxation from Iβ to IIII structure along the cellulose fibril main axis (direction perpendicular to the plane within which a0 , b0 , and γ0 are calculated). Partial structural relaxation along the cellulose fibril main axis is observed only in the MD simulation at very high temperature (T = 650 K) suggesting that longer simulation times may be necessary to obtain the complete fibril structural relaxation. The same set of simulations (#2) also provide microscopic details on the penetration of NH3 molecules into cellulose fibril. Recent diffraction and MD simulation studies9,10 were used to 9785

dx.doi.org/10.1021/jp2048844 |J. Phys. Chem. B 2011, 115, 9782–9788

The Journal of Physical Chemistry B

ARTICLE

Figure 3. Representative plots of the early (100 ps) time evolution of the tg, gt, and gg fraction populations for solventexposed chains at T = 413 and 298 K (A and B, respectively) as well as for cellulose chains in the crystalline core at T = 413 and 298 K (C and D, respectively). The plots are generated from values averaged over 20 trajectories. The errors vary between 3% and 6%.

estimate the maximum (ideal) number of ammonia molecules (222) which can be “placed” into our model cellulose fibril. The time series for the ammonia penetration in Figure 4 monitor the fraction of this number. In terms of the fraction of ammonia molecules within the fibril (NH3 fraction), the averaged data from the simulations at T = 413 K (Figure 4, panels C and D) show a much faster ammonia penetration with respect to the simulations at T = 298 K (Figure 4, panels A and B). In detail, the plots in Figure 4, panels C and D show a clear interplay between ammonia penetration and fibril structural changes, with partial relaxation for both a0 and γ0 (toward values compatible with the cellulose IIII structure) reached with very low ammonia penetration (NH3 fraction ∼0.2). The corresponding penetration of ammonia molecules occurs through the (100) and (100) surfaces of the fibril. In the new crystalline structure (see the bottom of Figure 1), the cellulose chains “sit” on top of each other forming stacks of cellulose molecules and “vertical channels” are created between the stacks. Our results suggest that the presence of these “channels” is necessary for the ammonia molecules to penetrate and eventually saturate the fibril. To confirm that, we ran an additional MD simulation (60 ns long) on a cellulose IIII fibril solvated with ammonia where the “vertical channels” between the stacks are present from the beginning of the simulation. Indeed, we observe that the ammonia penetration starts after a few simulation steps and proceeds rapidly to completely pervade

the fibril (data not shown). It is also worth noting that, the penetration of the ammonia molecules does not seem to affect the structure of our model cellulose IIII fibril. The analysis of the ammonia penetration up to a time scale of 50  70 ns shows that, after a first rapid growth, the NH3 fraction slowly increases up to ∼0.35. Additional MD simulations at T = 450, 550, and 650 K qualitatively suggest that the rate of the ammonia penetration process monotonically increases with temperature. In particular, after 30 ns at T = 650 K, we observe that NH3 fraction >99% and a quite distorted fibril loosely retaining a cellulose IIII structure (see Figure S5 in the Supporting Information). The results presented in this computational study provide clues on improving the current lignocellulosic biomass chemical pretreatments such as liquid ammonia and ionic liquids (ILs) pretreatments8,24 that rely on the disruption/alteration of the internal cellulose hydrogen bond network. First, our temperature dependent studies show that the liquid ammonia pretreatment should benefit from higher temperatures. This observation is consistent with a limited amount of information from the literature on liquid ammonia pretreatments25,26 and the role of thermal cycling on cellulose crystallinity.27,28 Importantly, we propose that the biggest gain at high temperature is seen in the eventual saturation of NH3 molecules into the cellulose fibril. The early events associated with the perturbation of cellulose surface and the formation of channels can occur very rapidly even at low temperature. Therefore one does not need high 9786

dx.doi.org/10.1021/jp2048844 |J. Phys. Chem. B 2011, 115, 9782–9788

The Journal of Physical Chemistry B

ARTICLE

Figure 4. Time evolution for the structural parameters a0 , γ0 and for the fraction of ammonia molecules within the cellulose fibril. (A and B) T = 298 K. a0 and γ0 , respectively. (C and D) T = 413 K. a0 and γ0 , respectively. The largest errors are 0.25 and 0.3 Å for a0 , at T = 413 and 298 K, respectively, and 3 and 12 for γ0 , at T = 413 and 298 K, respectively. For the fraction of NH3 within the fibril the largest errors are 0.03 and 0.004 for a0 , at T = 413 and 298 K, respectively. The plots show data averaged over 10 different trajectories for both T = 298 and 413 K simulations.

temperatures to perturb the cellulose structure29,30 as it can be done easily with bulk liquid ammonia at room temperature. One possible approach to improve existing ammoniabased pretreatments would be to use a gradient heating protocol to further decrystallize cellulose. Finally, our study shows that the ability of a chemical species to strongly interact with the cellulose fibril surface appears to be, at least, as important as its ability to penetrate into the fibril. Therefore, experimental protocols can be further refined by choosing chemicals that show specificity to the hydrophobic (100) and (100) fibril surfaces and at the same time exert enough chemistry to form an extended network of hydrogen bonds with the surface cellulose chains. With respect to ILs pretreatment, improvements may be obtained by tailormaking appropriate ILs that show high specificity for certain cellulose surfaces by polar anions without interference from the non polar cationic component (e.g., guanidium versus imidazolium or pyrimidium based cations).31 Our studies imply that some of the inefficiencies in liquid ammonia treatment may arise due to inaccessibility of the (100) and (100) surfaces. Indeed, the hydrophobic nature of these surfaces could facilitate lignin association and possibly impede cellulose accessibility to ammonia or ILs. One possible approach to improve the efficiency of liquid ammonia or ILs pretreatments will be to combine the pretreatments with carefully chosen organic solvents that can

selectively extract lignin (and/or hemicellulose) from these surfaces and enable access to ammonia or ILs.

’ CONCLUSION In summary, we provide a mechanistic model for early events in the liquid ammonia pretreatment of cellulosic biomass. The rapid formation of an extended hydrogen bond network between liquid ammonia and the cellulose molecules on the surface of the fibril induces a major shift in the hydroxymethyl group rotational state population. The change from tg to the gt and gg rotational states drives the structural crossover from cellulose Iβ to a new crystalline structure compatible with both the ammoniacellulose I complex and cellulose IIII. In the new crystalline structure, the cellulose chains stack on top of each other to form channels orthogonal to the (100) and (100) surfaces of the fibril. These channels allow ammonia molecules to penetrate into the cellulose fibril. Such a mechanistic model with the proposed early microscopic events is consistent with X-ray time-resolved studies of EDA penetration into cellulose microfibrils from Cladophora,32 thus suggesting that the mechanistic model deduced here for a small fibril can be applicable to larger fibrils as well. Finally, we have discussed possible ways in which the findings from the current studies can be incorporated into existing chemical pretreatments to make them more effective and economical. 9787

dx.doi.org/10.1021/jp2048844 |J. Phys. Chem. B 2011, 115, 9782–9788

The Journal of Physical Chemistry B

’ ASSOCIATED CONTENT

bS

Supporting Information. Additional figures referenced in the main manuscript. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]. Present Addresses r

Oak Ridge National Laboratory, Oak Ridge, Tennessee 378316475, United States.

’ ACKNOWLEDGMENT This work was funded by the LANL LDRDX98U program and the computational resources were provided by LANL institutional computing. S.P.S.C. and B.E.D. thank the DOE Great Lakes Bioenergy Research Center (DOE BER Office of Science DE-FC02-07ER64494) for support

ARTICLE

(21) Reddy, C. K.; Das, A.; Jayaram, B. J. Mol. Biol. 2001, 314, 619–632. (22) Mazeau, K.; Heux, L. J. Phys. Chem. B 2003, 107, 2394–2403. (23) Shen, T.; Langan, P.; French, A. D.; Johnson, G. P.; Gnanakaran, S. J. Am. Chem. Soc. 2009, 131, 14786–14794. (24) Liu, H.; Sale, K. L.; Holmes, B. M.; Simmons, B. A.; Seema, S. J. Phys. Chem. B 2010, 114, 4293–4301. (25) Yatsu, L.; Calamari, T.; Benerito, R. Text. Res. J. 1986, 56, 419– 424. (26) Lewin, M.; Roldan, L. G. J. Polym. Sci. Part C: Polym. Symp. 1971, 36, 213–229. (27) Cuculo, J.; Smith, C.; Sangwatanaroj, U.; Stejskal, E.; Sankar, S. J. Polym. Sci. Part A: Polym. Chem. 1994, 32, 229–239. (28) Cuculo, J.; Smith, C.; Sangwatanaroj, U.; Stejskal, E.; Sankar, S. J. Polym. Sci. Part A: Polym. Chem. 1994, 32, 241–247. (29) Wada, M. J. Polym. Sci., Part B 2002, 40, 1095–1102.  (30) Matthews, J.; Bergenstrahle, M.; Beckham, G. T.; Himmel, M.; Nimlos, M.; Brady, J.; Crowley, M. J. Phys. Chem. B 2011, 115, 2155– 2166. (31) Kahlen, J.; Masuch, K.; Leonhard, K. Green Chem. 2010, 12, 2172–2181. (32) Nishiyama, Y.; Wada, M.; Hanson, B.; Langan, P. Cellulose 2010, 17, 735–745.

’ REFERENCES (1) Ding, S.-Y.; Himmel, M. E. J. Agric. Food Chem. 2006, 54, 597–606. (2) Himmel, M. E.; Ding, S.-Y.; Johnson, D. K.; Adney, W. S.; Nimlos, M. R.; Brady, J. W.; Foust, T. D. Science 2007, 315, 804–807. (3) Binder, J. B.; Raines, R. T. Proc. Natl. Acad. Sci. U.S.A. 2010, 107, 4516–4521. (4) Shen, T.; Gnanakaran, S. Biophys. J. 2009, 96, 3032–3040. (5) Nishiyama, Y.; Langan, P.; Chanzy, H. J. Am. Chem. Soc. 2002, 124, 9074–9082. (6) Wada, M.; Chanzy, H.; Nishiyama, Y.; Langan, P. Macromolecules 2004, 37, 8548–8555. (7) Chundawat, S. P.; Bellesia, G.; Uppugundla, N.; da Costa Sousa, L.; Gao, D.; Cheh, A.; Agarwal, U. P.; Bianchetti, C. M.; Phillips, G. N. J.; Langan, P.; Balan, V.; Gnanakaran, S.; Dale, B. E. J. Am. Chem. Soc. 2011, 133, 1116311174. (8) Chundawat, S. P. S.; Beckham, G. T.; Himmel, M.; Dale, B. E. Annu. Rev. Chem. Biomol. Eng. 2011, 2, 121–145. (9) Wada, M.; Nishiyama, Y.; Langan, P. Macromolecules 2006, 39, 2947–2952. (10) Wada, M.; Nishiyama, Y.; Bellesia, G.; Forsyth, T.; Gnanakaran, S.; Langan, P. Cellulose 2011, 18, 191–206. (11) Phillips, J.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid, E.; Villa, E.; Chipot, C.; Skeel, R.; Kale, L.; Schulten, K. J. Comput. Chem. 2005, 26, 1781–1802. (12) Kirschner, K.; Yongye, A.; Tschampel, S.; Gonzalez-Outeiri~ no, J.; Daniels, C.; Foley, L.; Woods, R. J. Comput. Chem. 2008, 29, 622–655. (13) Case, D.; Cheatham, T.; Darden, T.; Gohike, H.; Luo, R.; Merz, K.; Onufriev, A.; Simmerling, C.; Wang, B.; Woods, R. J. Comput. Chem. 2005, 26, 1668–1688. (14) Gao, J.; Xia, X.; George, T. F. J. Phys. Chem. 1993, 97, 9241– 9247. (15) Feller, S.; Zhang, Y.; Pastor, R.; Brooks, B. J. Chem. Phys. 1995, 103, 4613. (16) Martyna, G. J.; Tobias, D. J.; L, K. M. J. Chem. Phys. 1994, 101, 4177. (17) Darden, T.; York, D.; Pedersen, L. J. Chem. Phys. 1993, 98, 10089. (18) Ryckaert, J.; Ciccotti, G.; Berendsen, H. J. Comput. Phys. 1977, 23, 327–341. (19) Mills, J. E.; Dean, P. M. J. Comput. Aided Mol. Des. 1996, 10, 607–622. (20) Luzar, A.; Chandler, D. Nature 1996, 379, 55–57. 9788

dx.doi.org/10.1021/jp2048844 |J. Phys. Chem. B 2011, 115, 9782–9788